What tech stack does Agility Robotics use?

Agility's robot software runs on C++ and Python, with ROS 2 (Robot Operating System 2) as the middleware layer — confirmed by engineering job postings requiring ROS 2 proficiency and by the NVIDIA Isaac ROS ecosystem that Agility builds upon. For AI model training, the company uses NVIDIA Isaac Lab, an open-source modular reinforcement learning framework built on Isaac Sim, which enables high-fidelity physics-based training in simulated environments before deployment on physical hardware. NVIDIA Isaac Sim provides the photo-realistic simulation environment, with GPU-accelerated physics based on PhysX 5 and real-time ray tracing to produce synthetic training data at scale.

The Cosmos generative world foundation model platform (via NVIDIA) provides synthetic data generation capabilities that accelerate training data creation without requiring physical robot time. NVIDIA Omniverse and the NVIDIA Mega Blueprint extend this infrastructure to fleet-scale simulation, enabling digital twin representations of entire customer facilities — bins, shelves, conveyor layouts, AMR paths — used for deployment planning and continuous simulation-based retraining as Digit's capabilities evolve.

For cloud data infrastructure, Agility's data platform engineering roles specify AWS, GCP, or Azure cloud experience alongside proficiency in distributed data frameworks including Apache Spark, Kafka, Beam, and Flink — consistent with a company building telemetry pipelines from more than 100 deployed robots generating operational data across multiple customer sites. Python (with likely PyTorch or JAX for ML work) and JVM languages (Java/Scala for data pipelines) appear in job postings for data engineering roles.

Agility Arc, the company's proprietary fleet management cloud platform, is the primary customer-facing software product. It integrates with warehouse management systems (WMS), manufacturing execution systems (MES), and autonomous mobile robots (AMRs) — meaning Agility's integration surface touches SAP EWM, Oracle WMS, Blue Yonder, Manhattan Associates, and similar enterprise installations at customer sites, though specific certified WMS integration partners have not been publicly enumerated. The Arc platform also handles OTA software updates, autonomous charging coordination, and safety compliance monitoring for deployed Digit units.

For perception, Digit uses a sensor stack combining LiDAR and multiple cameras, processed onboard using NVIDIA AI acceleration hardware, enabling navigation in dynamic warehouse environments without requiring dedicated infrastructure (no special floor markings or fixed fiducials). The robot communicates fleet status and telemetry to Agility Arc over standard enterprise networking protocols.

Agility's CRM and outbound sales tooling are not publicly disclosed. Given the Magic Leap backgrounds of CEO Johnson, CBO Diez, and CLO Lang — and Magic Leap's Salesforce-centric enterprise GTM — familiarity with Salesforce is likely, though the stack has not been confirmed. Marketing technology is similarly undisclosed and appears minimal, appropriate for a company at the stage of managing a few dozen enterprise accounts.

The depth of Agility's NVIDIA integration creates both constraints and entry points for vendors. The NVIDIA relationship is structural, not superficial: NVentures invested in the Series C, making NVIDIA a financial stakeholder, and Isaac Sim, Isaac Lab, Cosmos, and Omniverse are all actively used across the physical AI pipeline. Vendors with NVIDIA-compatible tooling — computer vision SDKs, synthetic data platforms, NVIDIA-certified edge inference hardware, or Omniverse connectors — have a natural wedge because Agility's AI team is already fluent in that ecosystem.

For data and cloud infrastructure vendors, Agility's multi-cloud posture (AWS and GCP signals in job postings) suggests there is no single-cloud lock-in. Vendors with strong Spark, Kafka, or Flink integrations and robotics telemetry use cases should find an open conversation. The WMS integration surface is the major near-term commercial opportunity as Agility scales customer deployments: any ISV that has pre-built connectors for SAP EWM, Blue Yonder, or Manhattan Associates can offer reduced integration burden at customer sites and is solving a real scaling bottleneck for Agility's deployment team.

GTM and CRM tooling is likely being evaluated or re-evaluated as Agility's commercial team scales post-Series C and moves from a handful of pilot accounts to a growing enterprise portfolio. Sellers in that category should engage with Daniel Diez (CBO) and Jennifer Hunter (CFO/COO) as the primary budget owners for sales enablement and operational software. Simulation and synthetic data vendors should engage through Pras Velagapudi (CTO) and the Pittsburgh engineering team.